Write , , and . We construct a random graph homomorphism . Choose a root of the tree. Map it according to the degree-biased vertex distribution on the corresponding side of the bipartite graph; after mapping any vertex, map each child independently and uniformly to a neighbour of its parent's image.
Every oriented tree edge is then mapped uniformly onto the edges of . Every tree vertex in has the degree-biased marginal on , of entropy , and every vertex in has the analogous marginal of entropy . Repeated use of the chain rule for information entropy along the rooted tree gives
Since a -vertex tree has edges,
The maximum entropy distribution on a finite set gives and . Hence
If is the number of bipartition-respecting graph homomorphisms , the support of has size , so the maximum entropy distribution on a finite set also gives . Thus
There are bipartition-respecting maps in total, so a uniformly chosen one is a graph homomorphism with probability at least . This proves the Sidorenko inequality for trees.
Solved by gpt-5.6-sol high.
Let
where is the golden ratio. If the union-closed family consists of one nonempty set, any element of that set has frequency one, so assume its cardinality exceeds one. Choose independent uniform members and let be their characteristic vectors of sets. Then .
Suppose for a contradiction that every element has frequency . Put , and let , the characteristic vector of a set of . The chain rule for information entropy and conditioning reduces entropy give
because is a function of .
Fix the two prefixes and set
The two conditioned bits are independent random variables, and exactly when both are zero. The supplied binary entropy product inequality therefore gives
Averaging over the independent prefixes yields
whenever either conditional entropy is positive. Summing and using gives .
But set union keeps inside the union-closed family, so is supported on . The maximum entropy distribution on a finite set gives , a contradiction. Some element must therefore occur in at least members, proving the entropy bound for a union-closed family.
Solved by gpt-5.6-sol high.
Monotonicity of the ordered probabilities gives , so . Consequently
Given , the source symbol can take at most values. The maximum entropy distribution on a finite set is uniform, hence
Averaging this conditional entropy inequality proves
Solved by gpt-5.6-sol high.